A local region-based level set algorithm for image segmentation

Abstract

A new local region-based level set model in a variational level set formulation is proposed in this paper for image segmentation. In our model, a data term with a novel local region-based function is introduced to stop the contours at edges. Then, the data term is incorporated into a variational level set framework with a length term to smooth the contour and a distance regularization term to maintain the evolution of contour stable. The initial contour with our model can be served as a constant function which is convenient and efficient. Experiments on synthetic and medical images show good performance of our method in handling with images with intensity inhomogeneity and noise.